Supplement B: Plant Microfossils

Summary

This supplement describes the laboratory preparation, analysis, and quantitative analysis of plant microfossils recovered from the Bargny 1 deposits.

Laboratory methods

Sediment samples were ground, passed through a 250-micron sieve, and placed in a shaker overnight with Calgon solution (sodium hexametaphosphate) before having sands and clays separated by settling and centrifugation-decant cycles at 2500 rpm for 2 minutes. At this point, samples were spiked with Lycopodium spores and treated with 10% HCl in a 40º bath for 10 minutes. After centrifugation-decant cycles until pH neutral, the samples were separated by density using a solution of zinc bromide and 5% HCl with a specific gravity of 2.3 g/ml. The resulting residue was extracted in ethanol and transferred to glycerol for analysis.

This set of laboratory methods for extracting phytoliths employs standard steps for digesting terrestrial sediments and recovering plant microfossils, but omits the use of a strong oxidizing agent such as peroxide or nitric acid. Judging from the well-oxidized state of the sediments, futher degradation of the organic remains would only remove valuable information represented by organic microfossils (pollen microfossils, aeomebas, etc.), but also permits the use of Lycopodium spores to track laboratory errors and to calculate microfossil concentrations.

Table 1. Samples, weights, tracer spores, diagnostic phytoliths, total phytoliths

	depth	stratum	LYCO_Count	LYCO_n	TODAL_diag
BG1.020	20	I	19855	239	151
BG1.040	40	II	20848	212	64
BG1.060	60	II	20848	57	227
BG1.090	90	II	19855	100	16
BG1.100	100	III	14285	200	29
BG1.110	110	III	19855	360	203
BG1.150	150	III	14285	423	256
BG1.175	175	III	19855	165	210
BG1.195	195	III	14285	344	208
BG1.210	210	III	19855	20	215
BG1.220	220	IV	14285	134	223
BG1.230	230	IV	14285	244	67
BG1.240	240	IV	14285	287	200
BG1.255	255	V	19855	163	374
BG1.270	270	V	20848	20	226
BG1.280	280	V	20848	11	231
BG1.290	290	V	19855	66	301
BG1.295	295	V	19855	78	231
BG1.300	300	V	19855	214	290
BG1.305	305	V	19855	309	213
BG1.310	310	V	20848	258	21
BG1.315	315	V	20848	162	196
BG1.320	320	VI	20848	114	236
BG1.325	325	VI	19855	370	206
BG1.330	330	VI	19855	119	203
BG1.335	335	VI	19855	465	277
BG1.340	340	VI	19855	49	216
BG1.345	345	VI	19855	44	39
BG1.350	350	VI	19855	0	13
BG1.355	355	VI	19855	3	214
BG1.360	360	VI	19855	400	207

Microfossil identificaiton

Phytolith and pollen microfossils were identified using a binocular light microscope at 400x-1000x. Phytolith nomenclature and categories follow the International Code for Phytolith Nomenclature (ICPT 2019), but we tried specifically to create sample categories consistent with Bremond et al.’s (2005, 2008) assessment of phytoliths from surface samples across West Africa (Table 2).

Table 2. Phytolith classifications from ICPT (2019) compared to Bremond et al. (2005)

	AFFILIATION	TYPE	GROUP	Bremond_et_al_2005
9	DICOTYLEDONS	SPH_ORN	WOODY PLANTS	Rough spherical
10	DICOTYLEDONS	SPH_CMP	WOODY PLANTS	Non-diagnostic
11	ARECACEAE	SPH_ECH	PALMS	Crenate spherical
12	MONOCOTYLEDONS*	ACU_BUL	MONOCOTS	Point-shaped
13	POACEAE_CYPERACEAE	BUL_FLA	MONOCOTS	Fan-shaped
14	POACEAE	PAP	GRASSES	Cone-shaped
15	POACEAE	RON	GRASSES	Non-diagnostic
16	POACEAE	BIL	GRASSES	Dumbell
17	POACEAE	CRO	GRASSES_PANICOID	Cross
18	POACEAE	SAD	GRASSES_CHLORIDOID	Saddle
19	POACEAE*	ELO_SIN	ELONGATE_GRASSES	Non-diagnostic
20	POACEAE*	ELO_DET	ELONGATE_GRASSES	Non-diagnostic
21	POACEAE*	ELO_DEN	ELONGATE_GRASSES	Non-diagnostic
22	VARIOUS	SPH_PSI	UNDIFF	Smooth spherical
23	VARIOUS	ELO_ENT	UNDIFF	Non-diagnostic
27	VARIOUS	BLO	UNDIFF	Fan-shaped
28	VARIOUS	UNDIFF	CORE_DETAILS	Non-diagnostic

Microfossil Images

Plant Microfossil Recovery, Preservation, and Concentration

Phytolith frequencies were quantified by counting all phytoliths until a sample size of 200 diagnostic phytoliths was reached (see Bremond et al. 2005, 2008). Concentrations of phytoliths and pollen were established by tracking Lycopodium spores encountered during analysis. Samples where 200 Lycopodium were encountered before 50 identifiable phytoliths were considered unreliable. Although the number of diagnostic pollen sampled does reach 100 in some samples, the number of damaged, fragmented, and indeterminate types is too high to pursue reliable statistical reconstructions with the pollen data. In general, the patterns in microfossil deposition/preservation closely follow the site’s sedimentology.

Figure 1. Phytolith and Pollen Recovery and Preservation

Phytolith Results

Phytoliths from the Bargny 1 samples show a strong representation of Chloridoideae and Panicoideae phytolith types (BIL and CRO, respectively) as well as types belonging to undifferentiated monocots (ACU_BUL, BUL_FLA, and RON). The frequency of SPH_ECH phytoliths is high in some of the lowermost samples, but we must use caution in interpreting this morphotype as it overlaps with sponge spicules common in coastal zones (ICPN 2009). While this type was tracked during analysis, this type is excluded from the statistical analyses. Samples with excessive SPH_PSI and SPH_ECH tend to be the most weathered and have the poorest preservation. These are excluded from futher analysis.

Figure 2. Phytolith results as a percent of the identifiable phytolith sum.

Pollen Results

Identifiable pollen was preserved in the samples and provides some insights into the kinds of vegetation cover that grew at the site during the major depositional phases. The lowermost samples are characterized by smaller pollen counts and low concentrations which steadily rise up to about 250 cm, where the samples yielded high pollen counts and a greater concentration of pollen. For much of the middle section of the deposits, the pollen are often damaged and although identified types are high, these tend to be types that are easily distinguished (Poaceae/Amaranthaceae). There’s a modest improvement in pollen preservation/concentration between 60-90 cm as well. What we do see in these samples are examples of types introduced by trade-wind activity (Olea-type & Pinus), riparian/estuarine trees/shrubs (Avicennia, Syzygium), and regional Sudanian woodland pollen types (Acacia, Celtis, Trema).

Figure 3. Pollen results as a percent of the identifiable phytolith sum.

Summary Plots by Plant Type

Figure 4. Phytolith and pollen percent by vegetation type.

Quantitative Comparisons with Surface Samples

Studies of surface soil phytoliths conducted by Bremond and colleagues (2005, 2008) provide an important resource for evaluating archaeological phytolith aseemblages. We use Principal Component Analysis and Minimum Square-Chord Distances to compare the phytolith spectra and to gain insights into the range of environments represented at Bargny 1. PCA was chosen because of the “predict” functionality in R, which is not available for other types of factor analysis (constrained correspondence, etc.). To look for changes in rainfall or the boundaries of major vegetation zones, we applied MSCD to establish the most similar sets of modern samples and plot their relevant climatic (rainfall) or geographic (latitude) representation across the depositional sequence.

Figure 6. PCA of modern soil sample (Bremond et al. 2005; 2008 - left) and Bargny 1 results plotted using the same eigenvalues.

PCA shows the influence of rainfall on Bremond et al.s (2005, 2008) phytolith spectra. However, the individual vegetation zones are only generally resolved using these vectors. Ellipses (65% and 95%) drawn around the important vegetation types near Bargny (Sahelian, Sudanian, and Saharan) to show the degree of overlap. The archaeological samples, when plotting using the eigenvectors from the original PCA, show a tight clusering with positive values on component 1 and low scores on component 2. Most of the samples fall within the 95% ellipse for Sahelian woodlands, but stratum 3 and 1 both yielded more negative component 1 loadings, indicating drier conditions.

Figure 7. Plots comparing Bargny samples with surface samples (Bremond et al. 2005) using square-chord distances of scaled (z-score) diagnostic phytolith results. The five best matches are plotted by their current precipitation regime per sampled depth.

Phytolith Indices for Savanna Types and Grass Stress

We can also compare the values for indices used by Bremond et al. (2005, 2008) to evaluate aridity/savanna type (Iph) and grass water stress (Fs). These values do a good job of discriminating the boundaries between the Sudanian and Guinean zones (Iph values) and the Sahelian and Saharan zones (Fs). These values allow us to look for potential turnover betwen major vegetation formations.

Figure 4. Comparison of long vs. short grass savanna index (Iph, x-axis) and Fan-shaped index (Fs, y-axis) results for Bargny samples (green and blue squares) with Bremond et al.’s (2005) results from West African surface samples

Synthetic Results

To create some insights into the scale and direction of past vegetation change represented in the Bargny 1 deposits, we plot the results for four different indices, grouping their results by stratum (Figure 6.). The indices include Fs, Iph, Grass-Amaranthaceae pollen, and wetland/riparian/mangrove pollen. As expected, the Grass/Amaranthaceae ratio is not especially informative, but the other indices show a clear turnover after stratum 4, when the depositonal environment shifts from near-coastal estuarine conditions to halophytic dry coastal plain.

Figure 6. Boxplots of Phytolith indices and grouped pollen results by stratum.

We further synthesize the results by plotting the Fs, Iph, and estuarine indices by stratum showing both the distribution of sample results (points) and their collective behavior (kernel density plots/color scheme).

Figure 7. Synthetic results showing wetland, water stress, and precipitation proxies with kernel density plots for each stratum